Inlining Research Articles

Inline expansion of SETL procedures

We report on an experiment in doing inline expansion of SETL procedures using the prototype of a preprocessor written in Franz LISP. This expansion facilitates in particular data flow analysis and enhances the quality of information which is available to the typefinding phase of the SETL optimizer.

Procedures and comments vs. the banker's algorithm

A 2 × 2 factorial experiment was performed to compare the effects of modularity (using internal procedures) and comments on the readability of the Banker's Algorithm. The readability of 4 editions of the program, each having a different combination of these factors, was inferred from the accuracy with which students could answer questions about the program after reading its text. Results of the experiment suggest that the Banker's Algorithm is more readable with in-line code than with internal procedures, and that the author's comments improve its readability.

Inline routines in VAXELN Pascal

This paper describes the implementation of inline procedures and functions inthe VAXELN Pascal compiler. Inline expansion translates the reverse Polish text produced by the parser into an intermediate language like that used in the VAX-11 PL/I and C compilers. The paper gives detailed descriptions of the front end's organization as it relates to inline routines and of the symbol substitutions made during inline expansion. The paper also discusses global optimization and the compiler's treatment of dynamically sized data, which is based on parametric types.

Software for Eight-Bit Microprocessors in Multivariable Digital Control Systems

This paper compares computer programs implementing linear multivariable digital control algorithms written in high level and assembly languages for eight-bit microprocessors. The execution speeds of the programs are compared for fixed-point and floating-point formats for controllers of increasing order, with faster speeds allowing greater achievable controller sampling rates. Methods are presented for optimizing the high level language programs using in-line assembly code in critical program modules and performing floating-point computations on a dedicated arithmetic function processor. Quantitative results are presented for achievable sampling rates as functions of the order of the controller, the number of inputs and outputs, language level used for the control program, and optimization method. Design guidelines are given for development of controller software for different applications.

Interfacing real-time operating systems to process control languages

The paper gives an overview over an operating system which provides for a flexible interface to support a wide spectrum of real-time language facilities. Due to its modular structure it can be tailored to the requirements of the language. In many cases this is achieved by an automatic, dialogue controlled generation of the system. For more serious changes well defined “basic functions” of the system have to be changed or new functions can easily be introduced. On the assembler level the system requests (SVC) are available as macro statements defined similar to real-time language syntax. A list of these macro calls is explicitly presented to give an overview over an implementation for the pdp 11 computer family; it shows the range in the definition of such a macro library. Based on this lowest level three methods to implement system requests in real-time languages are discussed. If the language allows assembler insertions the macros can immediately be used. If the real-time facilities of the language are connected via subroutine calls, they are realized by adding code procedures which contain the corresponding macros. In the third case, real-time operations are part of the syntax and can be compiled using the macros as inline code or generating adequate subroutine calls. These three methods are discussed with respect to readability, portability, software-safety, runtime efficiency and implementation costs. Implementations are illustrated by a few examples.

Predicting the effects of optimization on a procedure body

Practical application of procedure integration (inline expansion) as a program optimization requires some method for estimating the effects of subsequent optimization on the integrated procedure. A technique is described for predicting the code improvement that can be expected due to constant folding and test elision when a procedure call involving constant actual parameters is integrated. The technique is based on information collected during a single data flow analysis of each procedure body, and on execution frequency statistics for the procedure. This information can then be used to estimate the cost in code size and the benefit in execution speed of integrating each call to the procedure. The algorithm uses a syntax directed analysis technique suitable for structured programming languages without gotos.

Interfacing Real-time Operating Systems to Process Control Languages

The paper gives an overview over an operating system which provides for a flexible interface to support a wide spectrum of real-time language facilities. Due to its modular structure it can be tailored to the requirements of the language. In many cases this is achieved by an automatic, dialogue controlled generation of the system. For more serious changes well defined “basic functions” of the system have to be changed or new functions can easily be introduced. On the assembler level the system requests (SVC) are available as macro statements defined similar to real-time language syntax. A list of these macro calls is explicitly presented to give an overview over an implementation for the pdp 11 computer family; it shows the range in the definition of such a macro library. Based on this lowest level three methods to implement system requests in real-time languages are discussed. If the language allows assembler insertions the macros can immediately be used. If the real-time facilities of the language are connected via subroutine calls, they are realized by adding code procedures which contain the corresponding macros. In the third case, real-time operations are part of the syntax and can be compiled using the macros as inline code or generating adequate subroutine calls. These three methods are discussed with respect to readability, portability, software-safety, runtime efficiency and implementation costs. Implementations are illustrated by a few examples.

An analysis of inline substitution for a structured programming language

An optimization technique known as inline substitution is analyzed. The optimization consists of replacing a procedure invocation by a modified copy of the procedure body. The general problem of using inline substitution to minimize execution time subject to size constraints is formulated, and an approximate algorithmic solution is proposed. The algorithm depends on run-time statistics about the program to be optimized. Preliminary results for the CLU structured programming language indicate that, in programs with a low degree of recursion, over 90 percent of all procedure calls can be eliminated, with little increase in the size of compiled code and a small savings in execution time. Other conclusions based on these results are also presented.

Local optimizations

Some local optimizations (as opposed to global optimizations} are presented and a suitable one-pass compiler design for using them is shown. Optimizations shown are divided into machine dependent and independent classes with examples of each. There is some discussion of a subscript calculation technique which is an improvement over the usual technique, and a discussion of the best way to raise a quantity to a known small constant power by inline code. Various register allocation criteria are mentioned.